Anyone with a Clevo P950HR (or similar) with GTX1070 MaxQ tried the new 400+ nvidia drivers?

I've been having an issue with all the 400+ drivers where it seems to put a constant 5-7% usage on the "System" process (which normally hovers around 0-0.5% and peaks at 1% occassionally). While this may not seem like a lot, my battery life which is usually around 3hrs is dramatically reduced to under 1 hour. I reverted to the 399.24 drivers and it's all fine again.

Any ideas?

 

Have you done a clean install of the Intel + nvidia drivers?

 

Yeah, still get the same issue.

 

Have you reported it to nvidia?

 

Is the 399.24 driver being used from Clevo or from Nvidia. In case that sounds a little weird was the 399.24 driver supplied from Clevo?

 

From nvidia. Digging a little deeper: the nvidia GPU never goes to sleep after a monitor has been plugged into the displayport. Normally after it's unplugged you'd expect the GPU to go back to sleep. But after plugging in a monitor and then unplugging it, this doesn't seem to happen (and this kills battery life since the GPU is always on). So I wrote an app that monitors the power status. If you unplug the power and go onto battery, the app will disable the GPU device (same as disabling in device manager) and re-enable it. This fixed the problem of the GPU always being active. But for some reason with the 400+ driver version, this act of disabling and re-enabling in device manager causes the GPU to go crazy. Upon re-enabling, instead of staying asleep like on 399.24, it comes back to life with even more violence. Constantly stays open, and makes "System" process stuck on 6-7% CPU usage. 1hr battery life instead of 3. I'm just gonna stay on 399.24 forever.

 

Try the latest set.

 

That is more than likely causing the issue. I had a hard time with something similar with my Gtx 860's. Essentially at some point the newer driver affects the functionality of the device unless that driver update comes from the manufacturer. When I consulted with Nvidia they pointed out that drivers from the website vs oem are different animals. If I update past 388.71 I get black screens, reboots, hanging etc. issues that would indicate some sort of hardware problem. Reverting back to anything before 388.71 fixes the problem at the expense of lower frame rates.

 

I always use Nvidia drivers myself, they should work fine.

 

I wish I had your luck, but after fighting with it for quite some time I had to accept it. The nvldkmm.sys issue I encountered after upgrading drivers has a long history of individuals with the same problems eventually reverting to older drivers.

Since this thread I made below rolling back to the old drivers I have accumulated 100 hours of Rocket league, 30 hours of Dirt Rally, 40 hours of Titan fall 2, and 10 hours of Destiny 2 without a single issue.

http://forum.notebookreview.com/thr...ter-120hz-install.812183/page-4#post-10694492

 

It was also an issue on the MSI,

MSI AB and Steam webhelper becomes the cause of waking the dGPU on series 400+ driver.

You could try the fixed I posted here:
http://forum.notebookreview.com/posts/10812428/

The current version of the MSI Afterburner does keep waking the dGPU.

Fix is not coming any sooner stated by the author since he lacks hardware to test, you could do OC'ing using Nvidia Inspector on start up and use HWINFO64 to monitor.

Link here:
http://forum.notebookreview.com/thr...us-fix-doesnt-seem-to-come-any-sooner.825788/

 

Ah, interesting. I'm going to test this. I wanted to use Nvidia Inspecter for the OC as well as it's less intrusive. But for some reason it doesn't work with the app I wrote (that disables/reenables dGPU after DisplayPort and power are disconnected). Upon re-enabling, it also needs to re-enable the OC. By calling the OC shortcut for Nvidia Inspecter, it doesn't actually apply. But also killing and restarting MSI AB after re-enabling GPU did work, so I stuck with that.

 

Ok I just tried without MSI Afterburner (and I don't have Steam Web Helper)... still the same issue. I guess the disabling/reenabling of the device just has issues on its own with v400+.

 

I see, this method did work for me when I was digging through the problem.

Sometimes the NVIDIA defaults steam to dGPU, but if you don't use steam, maybe it is something else.

Also I added many of the possible exe to the windows graphic list, the one you saw in the guide, just to eliminate any other possibilities.

The way I found that the steam web helper is using the dGPU is through the task manager, there is a GPU usage column, maybe you can start from there.

I've been using NVIDIA Inspector's set OC command with Task Scheduler with a 45 seconds delay, counting starts once a user has logged in. And also set Windows 7 under compatibility mode. I believe it needs to wait for all the NVIDIA components to load before applying the OC, as I was getting memory not accessible error in the past, after doing this, all errors are eliminated.

I didn't force disabling / re-enabling the dGPU though, I just let it does what it do.

A lot of the modern laptops nowadays have the HDMI and other video output ports hooked up with the dGPU directly, if you are worrying about the battery life in video output mode, you can try some thunderbolt or USB to HDMI adapter which maybe more suitable in battery mode.

 

Ok, did a bit more testing. Switched my app to use Nvidia Inspector lnk to re-apply OC instead of MSIAB in a different manner, it's now working. But when I still saw the high System use and GPU staying awake, it was actually because GPU-Z was running (I had it open to see if the OC was re-applied or not). Closing GPU-Z makes everything OK again. So I guess your fix does work. It looks like certain GPU-monitoring apps cause this high CPU usage of "System", and MSI AB is probably one of them.

 

Funnily enough, without MSI AB now always running in the background using the new OC method and 416.94 drivers, I actually get even more battery life than the previous configuration, by about 30-40 minutes.

 

Glad it worked.

That list is actually the master list of switchable graphic, it seems to have a higher privilege than the NVIDIA panel. So it is basically assign once in the list, then you can leave it be, it'll still be there even you update the NVIDA driver. But the settings inside the NVIDIA panel is another story, I'll actually tweak in the NVIDIA panel again just to make sure both are correctly assigned.

If you need the NVIDIA panel to take over, you can set it to Default in the list, and the change regarding the switchable graphic within the NVIDIA panel shall apply.

 

What this guy above me said is what i've been having success with i'm on 418 at the moment and haven't had any issues but am noticing more and more apps want to use my nVidia instead of my Intel GPU which results in my GPU staying awake more (but not my CPU usage increasing) Also three parts of the Nvidia software now seem to want to keep things open. Anyway I was just going to suggest enable the Notification Icon in Nvidia Control Panel to keep an easier eye on what's using it, as windows only reports what's currently using it in foreground by the look of it, whereas the notification icon seems to be constantly monitoring? Don't know if that'll help but yeah I reckon every week i'm swapping the software defaults to Intel on more and more things.

 

This happens to me as well. I use an external monitor via miniDP for work and it's connected directly to Mains Power. Whenever there's a power outage, everything happily moves back on to the laptop screen however, the applications that was opened on the external monitor continue to run via the dGPU. You can see this list via the NVIDIA GPU Activity monitor. What I normally do is close all the applications shown on this list and re-open them. This causes the dGPU to go back to sleep. If not even my battery life is almost halved to about 1.5-2 hrs compared to iGPU 4-4.5hrs

 

This has always been the case for me, especially when using an external monitor. I created an app that runs in the background and monitors your power state and display count. If you switch to battery and have no external monitors connected, it will disable the 1070 device and re-enable it automatically 1 minute after removing the power. It can also be triggered by a hotkey on demand.

What this does is release the GPU and all apps using it, and when it's reenabled it stays "inactive" (does not consume any battery). Apps continue running just fine, 3D apps will obviously crash if they were actively using the GPU at the time that it was disabled. Browsers using GPU acceleration will be fine though but may require a page refresh.

Overall it works great for me because I can just unplug and move to the couch or something, and not worry about getting only 1hour of battery.

 

You cant switch the gpu a program us using on the fly and the outputs are driven by the dedicated gpu.

 

An usb-c / thunderbolt video adapter should get it back on iGPU as main. As the rendering is no longer directly coming from the dGPU.

 

IIRC the displayport over USB-c is d-GPU routed to support nvidia surround.